In our work we propose a deep learning solution to complete RGB-D images that are acquired by a NIR structured light scanner with an additional RGB camera that measures the visible spectrum. Building on works on image inpainting, we designed and trained a neural network architecture that takes the available fringe and color images as well as the reliably measured depth information and completes the depth images. We particularly focus on occlusion-caused image artifacts that naturally occur due to geometric visibility constraints. Hence, we are able to reconstruct a dense depth image from the viewpoint of the RGB camera, which can be used for further post-processing and visualization purposes.

Automatic analysis of image data is of high importance for many applications. Given an image classification problem one needs three things: (i) Training data and tools to extract (ii) relevant visual information—usually image features—that can be used by (iii) classification algorithms. For given (i), a multitude of candidates present themselves for (ii) and (iii). Model selection becomes the main issue. We present a web-based feature benchmark system enabling system designers to streamline tool-chains to specific needs using available implementations of candidate tools. Our system features a modular architecture, remote and parallel computing, extensibility and—from a user’s standpoint—platform independence due to its web-based nature. Using Wifbs, image features can be subjected to a sophisticated and unbiased model selection procedure to compose optimized pipelines for given image classification problems.

Visual images constitute one of the most common formats of electronic data that need to be efficiently accessed. In the context of this thesis, access can be understood as the retrieval of a specific subset of images from a larger set. The best method to identify this subset is problem dependent. At the heart of this thesis lies the quest for good design principles for the creation of semantic image retrieval systems. To state the problem more concisely:

The non-negative matrix factorization provides a valuable tool for the analysis of positive data by representing it as an additive linear superposition of a small number of non-negative basis elements. This property allows the base elements to be interpreted in the same domain as the input data. The problem though lies in the ambiguity of equally valid solutions from which only one is obtained. Its selection depends on the initialization of the applied factorization algorithm or further constraints.We propose a new approach which is based on sampling the set of valid factorizations, given one initial solution. First we derive a parameterization of the set of valid solutions by means of a strong membership oracle. This function returns true if a parameter tuple represents a valid solution and false otherwise. Furthermore, we present an algorithm that explores and samples parts of the non-convex solution set. To assist the otherwise automatic process and to alleviate the drawbacks of sampling a non-convex space, we provide a graphical user interface that puts the user in the loop. From an initial set of samples the user is allowed to select elements that serve as the starting point for subsequent samplings. With this browser-like tool a steering of the sampling of the NMF can be performed without further knowledge on the underlying algorithm and without the need to express possibly hard to formulate constraints. An evaluation of the sampling procedure reveals promising results for a factorization of data in up to 4 basis elements.

The usage of visual analytics during the analysis of business warehouse calculated key performance indicators is one emerging challenge in modern business applications. On the one hand, a complex network of key performance indicators has to be supervised. On the other hand, within this network only few key performance indicators change obviously within a short period of time. The sole mapping of the complexity of a network of key performance indicators to a graph-based visualization only covers static information and neglects temporal dependencies. We present a new visualization approach for the enrichment of graph-based visualizations of key performance indicator networks by introducing a multi-encoded visualization of additional functional, contextual and temporal information. The should help the user to understand relationships between KPIs and alert him if something is going wrong.

The bag-of-features model is often deployed in content-based image retrieval to measure image similarity. In cases where the visual appearance of semantically similar images differs largely, feature histograms mismatch and the model fails. We increase the robustness of feature histograms by automatically augmenting them with features of related images. We establish image relations by image web construction and adapt a label propagation scheme from the domain of semi-supervised learning for feature augmentation. While the benefit of feature augmentation has been shown before, our approach refrains from the use of semantic labels. Instead we show how to increase the performance of the bag-of-features model substantially on a completely unlabeled image corpus.

We present a novel approach for designing the search functionality in large unlabeled image databases. It combines Relevance Feedback, Hierarchical Browsing and Kernel PCA, uses a Mixture-of-Gaussian to model feature space distributions and different visualization techniques of high dimensional feature spaces. Given an image database, finding a specific single or set of pictures is achieved by assisting the user to find an as-short-as-possible browsing path through the database. Our system relies on describing each picture with an appropriate feature vector that results from applying Kernel PCA to image and textual based similarity matrices. We solve the page-zero-problem by presenting the centroids of a hierarchical clustering in feature space as initial suggestions. The user can then steer the search by selecting positive and negative examples which define a Mixture-of-Gaussian density in the parameter space. New suggestions are drawn according to this density and the user is thus directed to the desired image category. A user study proved our system to be practical and beneficial for category search tasks.

Amplitude spectra of natural images look surprisingly alike. Their shape is governed by the famous 1/f power law. In this work we propose a novel low parameter model for describing these spectra. The Sum-of-Superellipses conserves their common falloff behavior while simultaneously capturing the dimensions of variation--concavity, isotropy, slope, main orientation--in a small set of meaningful illustrative parameters. We demonstrate its general usefulness in standard computer vision tasks like scene recognition and image compression.

The browsing of large image data bases has become a standard problem not only on the web but also in private photo collections. Most browsing techniques build on high dimensional feature spaces that are reduced to one or two dimensions when presented to the user. As this approach does not scale well with the size of the data base we propose to use an interface based on the concept of parallel coordinates. Around the currently selected image, we collect images for each feature dimension, which vary only in this feature coordinate, and present them to the user in a row by row fashion. In this way the user can understand the individual feature dimensions independently. Besides directed image search the interface is well suited to explore classes of images and furthermore to evaluate how intuitive individual feature dimensions are for the user.

Stereoscopic movie production has been a topic in the making of films for a long time. However, it hasn't made it to the amateur sector. Commercially available stereoscopic cameras are too expensive for non professionals. When producing a stereo video with two separate standard cameras, synchronicity and spatial offset maintenance between the two views is a challenging non-trivial task. Even when this is done properly, the general lack of software-tools for stereo video post production definitely daunts any non-professional ambitions. In this work we present a tool for preprocessing stereoscopic videos. We describe how the input videos can be converted automatically to a stereo-movie, ready for displaying or further processing in standard video software.

This paper presents a new hardware-accelerated approach on volumetric reconstruction of trees from images, based on the methods introduced by Reche Martinez et. al [Rec04]. The shown system applies an adapted CT procedure that uses a set of intensity images with known interior and exterior camera parameters for creating a 3D model of a tree, while requiring considerably less images then standard CT. At the same time, the paper introduces a GPU-based solution for the system. As tomographic reconstructions are rather complex tasks, the generation of high-resolution volumes can result in very time-consuming processess. While the performance of CPUs grew in compliance with Moore’s law, GPU architectures showed a significant performance improvement in floating-point calculations. Regarding well parallelizeable processes, today’s end-user graphics-cards can easily outperform high-end CPUs. By improving and modifying the existing methods of volumetric reconstruction in a way, that allows a parallelized implementation on graphics-hardware, a considerable acceleration of the computation times is realized. The paper gives an overview over the single steps from the acquisition of the oriented images displaying the tree till the realization of the final system on graphics processing hardware.

The geometric referencing of digital image data and 3D point clouds e.g. given by a terrestrial laser scanner is the prerequisite for different levels of integrated data interpretation such as point cloud or mesh model texture colourisation for visualisation purposes, interactive object modelling by monoplotting-like procedures or automatic point cloud interpretation. Therein, the characteristics of laser scanner data and camera data can be regarded as complementary, so that these data are suitable for a combined interpretation. A precondition for the geometric referencing between laser scanner data and digital images and consequently for an integrated use of both data sets is the extraction of corresponding features. With a set of corresponding features the orientation of the image to the point cloud can be obtained by spatial resection. With regard to a high automation level the paper presents an approach for finding correspondences between features extracted from laser scanner data and digital images automatically. The basic idea of the presented approach is to use the SIFT operator to detect corresponding points in the camera image and an intensity image of the laser scanner data. Determining correspondences consists of four steps: Detection of salient characteristics, description of the features, matching of the descriptions in both images and evaluation of correct matches. RANSAC is used to find sets of consistent matches. The approach is validated with a data set taken from a baroque building in Dresden. * Corresponding author.

Untrained observers readily cluster paintings from different art periods into distinct groups according to their overall visual appearance or 'look' [WCF08]. These clusters are typically influenced by both the content of the paintings (e.g. portrait, landscape, still-life, etc.), and stylistic considerations (e.g. the 'flat' appearance of Gothic paintings, or the distinctive use of colour in Fauve works). Here we aim to identify a set of image measurements that can capture this 'naïve visual impression of art', and use these features to automatically cluster a database of images of paintings into appearance-based groups, much like an untrained observer. We combine a wide range of features from simple colour statistics, through mid-level spatial features to high-level properties, such as the output of face-detection algorithms, which are intended to correlate with semantic content. Together these features yield clusters of images that look similar to one another despite differences in historical period and content. In addition, we tested the performance of the feature library in several classification tasks yielding good results. Our work could be applied as a curatorial or research aid, and also provides insight into the image attributes that untrained subjects may attend to when judging works of art.

is a peer-reviewed scientific journal published exclusively in an electronic form. Its mission is to publish original contributions pertaining to the topics of Applied Computer Science, Information Systems and their Applications, to disseminate knowledge amongst its readers and to be a reference publication. The IADIS IJCSIS publishes original research papers and review papers, as well as auxiliary material such as short ongoing research papers, case studies, conference reports, management reports, book reviews and commentaries. This volume (Volume 9, Issue 2-ISSN: 1646-3692) combines 11 selected original papers that bring together researchers covering the wide spectrum of Theory and Practice in Modern Computing, Intelligent Information Systems Post-implementation and Change Management and their applications. The authors' contributions embrace significant research topics and intend to provide a current depiction of the research in the field while opening way to future research work. PERSONALISED SERVICE ROBOTS focuses on a service development strategy for a mobile social robot. The paper presents a service robot design based on the principles of a Service-Oriented Architecture (SOA), whose modularity design maximizes the advantages of multidisciplinary contributions from researchers of different areas. The key purpose of this research is the constant proactive provision of personalized support to elderly people, towards improving their quality of life and independence. The second paper, by Vesna Kirandziska and Nevena Ackovska, entitled A CONCEPT FOR BUILDING MORE HUMANLIKE SOCIAL ROBOTS AND THEIR ETHICAL CONSEQUENCE reports on the demands that occur in the human – robot interaction. The authors' purpose is to create social robots, in particular to develop more human-like empathic robots. Therefore, in order to better understand how human recognize other human's emotions a study about human-human emotion perception was done. Accordingly, to the authors the results " showed that humans are not so precise in perceiving other human's emotions. This result brought the idea of a new concept for making social emotion aware robots. " The third contribution by Alexander Streicher, Daniel Szentes and Wolfgang Roller with the title SCENARIO ASSISTANT FOR COMPLEX SYSTEM CONFIGURATIONS reflects on their concept and implementation of a mobile scenario assistant, which facilitates the automatic configuration of complex systems for demonstration scenarios. The authors present the features of a mobile assistance and e-learning system called Scenario Assistant (SCENAS), which conceals the difficulty of the underlying configuration of a complex, heterogeneous system for image exploitation. As a result, the addition of this work is the concept and implementation details …